name : vidhu sarwal
degree : cs '27 · thapar institute of engineering & technology
exp : samsung r&d · gsoc @ beagleboard · drdo
focus : llms · rag systems · distributed systems · embedded linux · ai infra
won : smart india hackathon 2024 (10k+ teams) · wro national top 3
status : building systems that don't break at scalelanguages
ai / ml
backend / infra
hardware / embedded
- distributed storage layer pooling multiple cloud providers (e.g. gdrive accounts)
- zero-knowledge design → server never stores keys, seed, or chunk map
- file sharding via ChaCha20-DRBG obfuscation across multiple chunking strategies
- secure token storage using AES-256-GCM encryption at rest
- handles 1000+ concurrent operations with sub-500ms latency
- converts any repo into a readable narrative
- AST-based parsing using Tree-sitter for function-level understanding
- LLM-powered summaries + commit evolution stories
- semantic search over 50k+ LOC using ChromaDB
- reduces onboarding time by ~60%
- extended hardware testing framework in C/C++
- mikroBUS Click auto-detection (I2C scan + EEPROM parsing)
- added HDMI validation, web dashboard, MQTT publishing, CI pipeline
- ~30% performance improvement
if it only works on your machine, it's not done.